arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2401.07575cs.CLcs.LGcs.SDeess.AS
收藏

用于音频-文本分类的级联跨模态 Transformer

Cascaded Cross-Modal Transformer for Audio-Textual Classification

National University of Science and Technology Politehnica Bucharest(布加勒斯特理工国立科技大学);University of Bucharest(布加勒斯特大学)

Nicolae-Catalin Ristea, Andrei Anghel, Radu Tudor Ionescu

AI总结:

针对训练数据有限的语音分类问题,提出级联跨模态 Transformer(CCMT),通过结合多语言 BERT 和 Wav2Vec2.0 音频特征进行多模态学习,在多项基准数据集上取得优异性能并获胜。

摘要:

语音分类任务通常需要强大的语言理解模型来掌握有用的特征,当可用训练数据有限时,这便成为问题。为了获得更优的分类性能,我们提出通过使用自动语音识别(ASR)模型转录语音,并利用预训练翻译模型将转录文本翻译成不同语言,从而利用多模态表示的内在价值。因此,我们为每个数据样本获得了音频-文本(多模态)表示。随后,我们通过一种新颖的级联跨模态 transformer(CCMT),将特定语言的 Bidirectional Encoder Representations from Transformers(BERT)与 Wav2Vec2.0 音频特征相结合。我们的模型基于两个级联的 transformer 块。第一个块结合来自不同语言的特定文本特征,而第二个块将声学特征与第一个 transformer 块先前学习到的多语言特征相结合。我们在 ACM Multimedia 2023 Computational Paralinguistics Challenge 的 Requests Sub-Challenge 中部署了我们的系统。CCMT 被宣布为获胜方案,分别在投诉和请求检测中获得了 65.41% 和 85.87% 的未加权平均召回率(UAR)。此外,我们将该框架应用于 Speech Commands v2 和 HarperValleyBank 对话数据集,超越了以往在这些基准上报告结果的研究。我们的代码可在以下网址免费下载:https://github.com/ristea/ccmt。

英文摘要:

Speech classification tasks often require powerful language understanding models to grasp useful features, which becomes problematic when limited training data is available. To attain superior classification performance, we propose to harness the inherent value of multimodal representations by transcribing speech using automatic speech recognition (ASR) models and translating the transcripts into different languages via pretrained translation models. We thus obtain an audio-textual (multimodal) representation for each data sample. Subsequently, we combine language-specific Bidirectional Encoder Representations from Transformers (BERT) with Wav2Vec2.0 audio features via a novel cascaded cross-modal transformer (CCMT). Our model is based on two cascaded transformer blocks. The first one combines text-specific features from distinct languages, while the second one combines acoustic features with multilingual features previously learned by the first transformer block. We employed our system in the Requests Sub-Challenge of the ACM Multimedia 2023 Computational Paralinguistics Challenge. CCMT was declared the winning solution, obtaining an unweighted average recall (UAR) of 65.41% and 85.87% for complaint and request detection, respectively. Moreover, we applied our framework on the Speech Commands v2 and HarperValleyBank dialog data sets, surpassing previous studies reporting results on these benchmarks. Our code is freely available for download at: https://github.com/ristea/ccmt.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑